Borrowing it
Nothing to install: this file belongs to sushiHex/hardline-mcp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/sushiHex/hardline-mcp/main/CLAUDE.mdgit clone --depth 1 https://github.com/sushiHex/hardline-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/sushihex/hardline-mcp/claude-md)<a href="https://agentmods.dev/instructions/sushihex/hardline-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/sushihex/hardline-mcp/claude-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/sushihex/hardline-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/sushihex/hardline-mcp/claude-md.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.04181 | $0.04181 |
| Opus 5.5 | $0.01672 | $0.01672 |
| Sonnet 5.5 | $0.00836 | $0.00836 |
| Haiku 4.5 | $0.00418 | $0.00418 |
Grade A, and why
hardline-mcp CLAUDE.md scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
last. `subprocess.run(text=True)` decodes as cp1252; `print()` encodes as How it starts
The opening of the file, as written. The whole thing — 300 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
hardline-mcp — working notes
The README explains what this is. This file is for working on it: the conventions, the gotchas that cost real time, and the rules the code already follows so new code follows them too.
Commands
python -m pip install -e ".[dev,codex-watch]" # codex-watch un-skips the Codex wake tests
python -m pytest tests/test_sessions.py -q # one module
python -m pytest "tests/test_identity.py::test_verified_owner_token_survives_later_probe_failure" -q
python -m pytest -q -rs -n auto # full suite, as CI runs it
python scripts/mutate.py [case ...] [--jobs N] # regression mutations (tests/mutations/*.json)
hardline-mcp report [--db PATH] # read-only audit of a store (report.py)
No linter or formatter is configured. CI (.github/workflows/ci.yml) runs three
jobs side by side: the suite on Ubuntu + Windows + macOS, Python 3.10 and 3.14, in
xdist workers, then an import of hardline_mcp.server to check that the tools
register; the mutation catalog once; and the live sandbox tests on Linux and
macOS. Don't run hardline-mcp by hand to test it. It is a stdio server and
blocks waiting for input.
Module map
Dependencies point one way: procid <- mailbox <- jobs/sessions <-
watch <- server.
cli.pypicks the subcommand (serve /watch/watch-codex/report) before importing anything heavy, so the watchers never loadmcporwebsocketsunless they need them.server.pyis the only module that importsmcp. Everything else is pure logic that takes an injectabledb_path/now_fn, so tests can drive it directly. Every tool isasync defand runs its blocking body through_in_thread(anyio.to_thread). A sync tool would block the event loop, pings included.mcpis pinned<2.0.0: 2.x removedmcp.server.fastmcp.mailbox.pyowns the SQLite connection, WAL, themessagestable, andSCHEMA_VERSION.jobs.py(jobstable) andsessions.py(agent_sessions) reuse its_connect/_resolve_db. All three tables are in one file on purpose, so a job's result and its inbox notice can commit in one transaction.procid.pyhandles liveness (ALIVE/DEAD/UNKNOWN) and creation-time process keys, Windows included.adapters.pyspawns the agent CLIs (hermes chat,codex exec,claude -p). It resolves the executables, sets sandbox/read controls, and decodes output.dispatch.pyis the cancellable executor behindask_*_async.github_context.pyis pure logic forgithub=: collect a PR through an injectedghrunner, store snapshots content-addressed on disk, and render what one reviewer receives.adapters.pysupplies the runner (under the call'son_spawn) and pipes the result to the child's stdin.watch.pyis a read-only, level-triggered observer of unread mail (what Claude's Monitor uses).wake_codex.pyforwards the same observations to a bound Codex app-server thread.channel.pydecides what to push into a Claude Code session and when (pure logic, no MCP import);delivery.pyrecords push facts and derives receipted/awaiting/unreceipted.server.serveserves throughserver.serve_streams- FastMCP's ownServer.runwithchannel.tapand the pusher spliced onto the raw streams - notmcp.run(). Seedocs/push-delivery.md.announce.pyis the notice wake beside the pusher in the same loop: one constant notice outstanding until its own receipt returns, new mail only. Two subclasses supply the address and transport:claude_inbox.pyposts into the Claude Code session's own cross-session inbox (the pipe and token in hardline's environment, proven the host's by~/.claude/sessions/<pid>.json). It replaces channel push whenever the inbox is proven; push is the fallback. Seedocs/claude-inbox-wake.md.codex_queue.pypins the Codex thread fromtools/call_meta(through the tap'son_callhook), steers a notice into the turn that last called hardline (codex_steer.py:turn/steeron the local daemon, overcodex app-server proxy), and queues withcodex queuewhen the daemon refuses. Seedocs/codex-queue-wake.md.
hints.pyremembers which Claude conversation last held each claimed name (lane_hints), so a relaunched or reconnected conversation is told about its old names and their mail. Advice only: it never claims, routes or reads. Seedocs/session-continuity.md.bindings.pyrecords which Codex thread each claimed name belongs to (thread_names), written only by attributed claims, restorations, releases and dismissals. A Codex server's wake poll claims its thread's names back through the ordinary ownership rule (sessions.survey/sessions.restore,server._restore_pending), and async admission refuses a result addressed to the parent-derived lane while the thread has names the server does not hold. See "A Codex thread's names" indocs/session-continuity.md.report.pyaudits a store read-only (mode=ro, one snapshot, nevermailbox._connect): which revisions the registered servers run (recorded inprocess_revisions, keyed by process identity), whether the legacysessionstable still has live writers, and who acked each job notice against who dispatched the job. Evidence, not verdicts: seedocs/architecture.md, "Who acked a message".
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 10d ago Changed · +2 lines · +37 tokens per session c75f3a5dc3e8
- 14d ago Changed · +23 lines · +402 tokens per session d4c41ece2696
- 16d ago Changed · +25 lines · +469 tokens per session d4f562cd9edb
- 17d ago Changed · +4 lines · +82 tokens per session 79225f4689dd
- 20d ago Changed · +20 lines · +364 tokens per session 9a2821b391a1
- 22d ago Changed · +5 lines · +89 tokens per session 372de32af2c1
- 23d ago Changed · +4 lines · +72 tokens per session 6420a3ed3f0c
- 27d ago Changed · +13 lines · +210 tokens per session ce413734d916
- 1mo ago Changed · +58 lines · +832 tokens per session af0a2d40f284
- 1mo ago First seen · 146 lines · 1,624 tokens per session scan A 05b95beb9fec
hardline-mcp CLAUDE.md is an instructions file published in the GitHub repository sushiHex/hardline-mcp (1 stars, last pushed 2d ago), licensed MIT. It adds 4,181 tokens to every session, about $0.0167 per session on Opus 5.5. A static security scan graded it A with 1 finding (runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-10.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deno copilot-instructions.md
Copilot instructions for denoland/deno, covering deno development guide for github copilot, network access, high level overview, key directories and key files to understand first.